Linus Torvalds on 5/13/07 wrote:
---------------------------
>[...]
>>So, in practical terms, is there benefit from using
>>64-bit OS on a machine with, lets say, 2-3 GB RAM ?
>
>Absolutely. No question what-so-ever. If you have 2GB
>of RAM on a 64-bit architecture, you can access it all
>easily in the kernel, and none of it is limited to just
>some special use. And a single big process can also use
>it effectively at the same time, so a database can
>actually have it all mapped without having to play
>windowing games in user space either.

Thanks. This was very informative. I guess I didn't actually realize that the HIGHMEM limit is not 4GB but 1GB-128MB (I guess depending on VMSPLIT_xx).

It appears that as far as kernel complexity is concerned, PAE (CONFIG_HIGHMEM64G) is not that much worse than simply having to support >1GB RAM (CONFIG_HIGHMEM4G).

If anybody is interested in some more details, I found this high level overview:
http://kerneltrap.org/node/2450

Mel Gorman's book "Understanding The Linux Virtual Memory Manager" covers it in some more detail. It is available online here:
http://www.phptr.com/content/images/0131453483/downloads/gorman_book.pdf

It seems the complication is already there and isn't ever going away. All driver code that needs to use the CPU to access process memory outside of the current process context needs to use kmap()/kunmap(), even though on a 64-bit system they don't do anything.

regards,
Tzvetan